{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:ETLA5BZ5L4YFKZDITB43K7YPQI","short_pith_number":"pith:ETLA5BZ5","schema_version":"1.0","canonical_sha256":"24d60e873d5f305564689879b57f0f822a9524a1fa76eb55e3246ab7a15e0def","source":{"kind":"arxiv","id":"2605.25114","version":1},"attestation_state":"computed","paper":{"title":"Counterfactually Safe Reinforcement Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"stat.ML","authors_text":"Chengchun Shi, Jingyi Li, Peng Wu","submitted_at":"2026-05-24T14:54:06Z","abstract_excerpt":"Reinforcement learning algorithms are generally designed to maximize the expected return across a population. However, a policy that is optimal on average may be suboptimal for certain individuals, leading to potential safety concerns. To address this, we first formalize the notion of individual harm from a counterfactual perspective and define harm as the event in which a chosen action results in a strictly worse outcome than a baseline alternative. We then propose a general two-stage procedure for learning policies that maximize the expected return while accounting for individual harm. We fu"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2605.25114","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"stat.ML","submitted_at":"2026-05-24T14:54:06Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"2f586a29025fc1d87683c9cb9b0c1496b73a483edd0a92c5fcef30dda20588a1","abstract_canon_sha256":"74083df583794cee177fac4f8b7583df321f7392f28885d0666397560b9f14b1"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-26T02:04:18.688329Z","signature_b64":"gmgVQpg0xaFlHrrOx5ceuJOkEc4scqFyNxVHgsdvwuXXkUm5tq3ne9Yb9bRsh5QhDs3EBSTAjs+EiCdKQR4wCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"24d60e873d5f305564689879b57f0f822a9524a1fa76eb55e3246ab7a15e0def","last_reissued_at":"2026-05-26T02:04:18.687532Z","signature_status":"signed_v1","first_computed_at":"2026-05-26T02:04:18.687532Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Counterfactually Safe Reinforcement Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"stat.ML","authors_text":"Chengchun Shi, Jingyi Li, Peng Wu","submitted_at":"2026-05-24T14:54:06Z","abstract_excerpt":"Reinforcement learning algorithms are generally designed to maximize the expected return across a population. However, a policy that is optimal on average may be suboptimal for certain individuals, leading to potential safety concerns. To address this, we first formalize the notion of individual harm from a counterfactual perspective and define harm as the event in which a chosen action results in a strictly worse outcome than a baseline alternative. We then propose a general two-stage procedure for learning policies that maximize the expected return while accounting for individual harm. We fu"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.25114","kind":"arxiv","version":1},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.25114/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"aliases":[{"alias_kind":"arxiv","alias_value":"2605.25114","created_at":"2026-05-26T02:04:18.687651+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.25114v1","created_at":"2026-05-26T02:04:18.687651+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.25114","created_at":"2026-05-26T02:04:18.687651+00:00"},{"alias_kind":"pith_short_12","alias_value":"ETLA5BZ5L4YF","created_at":"2026-05-26T02:04:18.687651+00:00"},{"alias_kind":"pith_short_16","alias_value":"ETLA5BZ5L4YFKZDI","created_at":"2026-05-26T02:04:18.687651+00:00"},{"alias_kind":"pith_short_8","alias_value":"ETLA5BZ5","created_at":"2026-05-26T02:04:18.687651+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/ETLA5BZ5L4YFKZDITB43K7YPQI","json":"https://pith.science/pith/ETLA5BZ5L4YFKZDITB43K7YPQI.json","graph_json":"https://pith.science/api/pith-number/ETLA5BZ5L4YFKZDITB43K7YPQI/graph.json","events_json":"https://pith.science/api/pith-number/ETLA5BZ5L4YFKZDITB43K7YPQI/events.json","paper":"https://pith.science/paper/ETLA5BZ5"},"agent_actions":{"view_html":"https://pith.science/pith/ETLA5BZ5L4YFKZDITB43K7YPQI","download_json":"https://pith.science/pith/ETLA5BZ5L4YFKZDITB43K7YPQI.json","view_paper":"https://pith.science/paper/ETLA5BZ5","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.25114&json=true","fetch_graph":"https://pith.science/api/pith-number/ETLA5BZ5L4YFKZDITB43K7YPQI/graph.json","fetch_events":"https://pith.science/api/pith-number/ETLA5BZ5L4YFKZDITB43K7YPQI/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/ETLA5BZ5L4YFKZDITB43K7YPQI/action/timestamp_anchor","attest_storage":"https://pith.science/pith/ETLA5BZ5L4YFKZDITB43K7YPQI/action/storage_attestation","attest_author":"https://pith.science/pith/ETLA5BZ5L4YFKZDITB43K7YPQI/action/author_attestation","sign_citation":"https://pith.science/pith/ETLA5BZ5L4YFKZDITB43K7YPQI/action/citation_signature","submit_replication":"https://pith.science/pith/ETLA5BZ5L4YFKZDITB43K7YPQI/action/replication_record"}},"created_at":"2026-05-26T02:04:18.687651+00:00","updated_at":"2026-05-26T02:04:18.687651+00:00"}